Discord thread starter 缓存加 TTL+LRU，防止长期内存膨胀（PR #5274）

• Discord integration is active in a busy server with frequent thread activity.
• You can observe gateway memory trend over at least 30-60 minutes.

1. Record baseline memory and cache-related behavior during thread-heavy traffic.
2. Upgrade to build containing PR #5274 and restart gateway.
3. Generate thread traffic continuously (create/reply/archive cycles) for a fixed test window.
4. Confirm stale cache entries expire around TTL and working set stabilizes near LRU cap.
5. Keep one regression check for thread-reply correctness after cache churn.

• Extremely bursty workloads can still cause short-lived peaks even with TTL/LRU controls.
• Exact memory impact depends on Node runtime and other active plugins（需验证）。

This tip is aggregated from community/public sources and preserved with attribution.